The valve on Maple Street was binding after eight months. I repacked it Tuesday morning, logged it as early maintenance, moved the next service date up to six months. The system would track it as accelerated wear, add it to the pattern I'd been documenting for eighteen months.
Valves failing at eight months instead of twenty-four. Relined sections breaking at three years instead of twenty. New installations showing stress at rates that didn't match their hundred-year ratings. Not catastrophic failures. Just steady compression of every timeline the schedule assumed.
The foreman asked if I wanted to flag the valve for replacement. I said yes. Not because the repacking wouldn't hold—it would, for now—but because the pattern suggested we were past the point where adjustments could keep pace.
By afternoon I was on Jefferson Street, where a relined section from three years ago had developed a circle break. The crew was cutting out the failed section. The liner had been rated for twenty years. It made it three.
I'd tracked fifteen relined sections over the past two years. Average lifespan: four years. The schedule assumed twenty. That wasn't adjustment territory anymore. The schedule's fundamental assumptions were wrong.
"Thought this one was good," the foreman said.
I figured we all did. But thinking something would hold and it actually holding were different things. The schedule was built on projections from infrastructure aging under historical conditions. Conditions weren't historical anymore. Heavier storms, more temperature swings, soil stress the old ratings didn't account for. The infrastructure was responding to something the schedule didn't recognize.
I'd been documenting it—not just in the official logs, but in a separate tracking sheet that showed the pattern across the whole system. Valve wear rates, liner failures, pressure fluctuations, break frequencies. All accelerating. All compressing timelines faster than the schedule could adjust.
Whether to respond wasn't the issue. We were responding—repacking valves early, monitoring sections closer, adjusting service dates. But incremental response couldn't match systemic acceleration. The data showed that clearly enough.
I drove past the Oakwood section on my way back—new ductile iron main from last year, rated for a hundred years. The old pipe they'd pulled out was from 1952, had failed at seventy years. The new pipe would probably fail sooner. Maybe at sixty years. Maybe fifty. The schedule assumed a hundred.
That night I finished the report I'd been working on for three months. Not the quarterly maintenance summary. A different report. One that laid out the tracking data, the pattern of accelerated failures, the compression of every timeline across the system.
The conclusion was straightforward: the schedule's assumptions about infrastructure lifespan were based on conditions that no longer applied. Adaptive maintenance could manage variation—some sections failing early, some lasting longer—but it couldn't manage systematic acceleration. When everything was failing faster, adjusting individual timelines just meant adjusting them again sooner.
The report recommended shifting from adaptive maintenance to emergency replacement protocols. Not wholesale overhaul—the money wasn't there, the disruption would be massive—but prioritized replacement of high-risk sections based on failure probability rather than scheduled service dates. Transparent communication with residents about actual system reliability rather than projected reliability. Revised lifespan assumptions for all new installations.
It meant acknowledging that the schedule we'd been following wasn't matching reality. That the adaptive approach, which had worked for decades, wasn't sufficient for conditions we were operating under now. That the timeline between repairs wasn't just compressing—it was compressing faster than we could adjust.
I didn't know if the report would change anything. The utility had invested heavily in the adaptive maintenance framework. The political will for emergency protocols wasn't there. The funding for accelerated replacement didn't exist. But the data was clear enough that someone needed to say it.
The valve on Maple Street would need service again in six months. The Jefferson Street replacement would hold for now. I'd check both next month, document what I found, add it to the pattern. The work was careful. The system kept running.
But someone needed to say that careful work and running systems weren't guarantees of reliable systems. That adjusting the schedule worked when the schedule's assumptions were basically sound. That when the assumptions were wrong, adjusting just meant managing decline more slowly.
I submitted the report Thursday morning. The valve on Maple Street would probably hold until the next service date.
Probably.